Data-Driven Fault Detection and Identification in Wind Turbines Through Performance Assessment

International audience This paper proposes a simple data-driven method for fault detection and identification based on the assessment of performance and detection miscorrelation among the operating data of wind turbines. The operating data were collected over one year on a 30 MW wind farm in Mauritania, Nouakchott area, for parameters including the wind speed, generator output power, turbine rotating speed, generator temperature, rotor speed, gearbox bearing temperature and generator rotating speed. The method starts with identifying the distribution of the pitch angle and the temperature of the generator on each of the four identified operating phases of the wind turbine. Then the correlation faults analysis allows identifying their characteristics and explaining the performance failures.